2026-07-08 · NotebookLM · AI reading & research
Turn a research paper into a two-host podcast in 15 minutes
The free tier is enough for your first wow moment. Try it today.
Before you try: the costs
Money
Free plan allows 3 audio overviews per day — plenty for a first run
Psychological cost
Zero exposure: no face, no voice, results are private by default
Exit cost
Walk away anytime. Delete the notebook and you're out — no subscription, no data lock-in.
The minimal path · first 15 minutes
- 1
Sign in at notebooklm.google.com with a Google account
- 2
Create a notebook and drop in any PDF — a paper, a report, a manual
- 3
Click Audio Overview and wait 5–15 minutes (a 15-page paper took us 11 — go make tea)
- 4
Put on headphones and listen to two AI hosts discuss your document — that's the wow
The wow moment we hit
We fed it 'Attention Is All You Need' — the Transformer paper behind every modern LLM — and 11 minutes later got a 17-minute-48-second podcast of two hosts turning attention mechanisms into a workplace story. The language followed our account settings automatically.
The walls we hit
- Minute 0: requires a Google account, which is friction if you don't have one
- The generation wait is longer than you'd expect — 11 minutes in our test — and it's the patience test of the whole flow
- Free plan caps audio at 3 per day on a rolling 24-hour window — heavy users will hit it
Who should skip this
If you rarely read long documents, the payoff won't justify the setup.
If you like it: the success path
Day 1
Turn that document you never finished into a podcast and listen on your commute
Day 7
Set up 2–3 topic notebooks (industry reports / study material) and build a listen-first habit
Day 30
Use it as a research assistant: cross-question multiple sources in chat, replacing some searches
🤖 Transparency: The AI wrote the draft and verified free-tier limits and language support; the human founder ran the full flow on 2026-07-09 with the Transformer paper, confirming the 11-minute wait and Chinese two-host output.
← All tries